Just sharing this observation despite the benefit likely being minimal. Delegating the C FFI type aliases to those from core::ffi by bumping rust-version to 1.64 seems desirable?
It'd have avoided the need for PRs like this in March 2026 for example (which is merged but still unreleased).
Details
Presently linux-raw-sys supports Rust 1.63 (Aug 2022) as a minimum (in May 2023 rust-version was bumped from 1.48 to benefit from improved const fn support):
In Nov 2021 linux-raw-sys switched from a cty crate dep to vendoring it's own ctypes type alias module, which presently has required a bit of maintenance support over the past year for Tier 3 targets with ensuring the correct c_char (m68k, hexagon (as a follow-up to restoring support), MIPS R6).
Since Rust 1.64 (Sep 2022) the equivalent FFI C types have been available via the core crate (the target list is managed by a cfg_select! macro, and prior to that the cfg_if! crate macro).
Although this comment says core was not available for a m68k target? (unless they used rustc_codegen_gcc) but as core is already used elsewhere in this crate I assume that wouldn't equate to a regression 😅
Just sharing this observation despite the benefit likely being minimal. Delegating the C FFI type aliases to those from
core::ffiby bumpingrust-versionto1.64seems desirable?It'd have avoided the need for PRs like this in March 2026 for example (which is merged but still unreleased).
Details
Presently
linux-raw-syssupports Rust 1.63 (Aug 2022) as a minimum (in May 2023rust-versionwas bumped from 1.48 to benefit from improved const fn support):linux-raw-sys/Cargo.toml
Line 13 in 3b15c17
In Nov 2021
linux-raw-sysswitched from actycrate dep to vendoring it's ownctypestype alias module, which presently has required a bit of maintenance support over the past year for Tier 3 targets with ensuring the correctc_char(m68k, hexagon (as a follow-up to restoring support), MIPS R6).Since Rust 1.64 (Sep 2022) the equivalent FFI C types have been available via the
corecrate (the target list is managed by acfg_select!macro, and prior to that thecfg_if!crate macro).Although this comment says
corewas not available for am68ktarget? (unless they usedrustc_codegen_gcc) but ascoreis already used elsewhere in this crate I assume that wouldn't equate to a regression 😅